Share with your CIO
Andover Companies, a 198-year-old Northeast mutual insurer, is betting that getting the data foundation right before deploying AI is the only sequencing that works. Under a five-year deal with Cognizant, the insurer will consolidate its policy administration platform integrations, build an AWS-native enterprise data layer using a Medallion architecture, harden its policyholder and agent portals for NYDFS cybersecurity compliance, and then, only then, begin prototyping generative and agentic AI for underwriting intake and claims triage. Financial terms were not disclosed.
What this means for your business
The insurer sitting across from you at renewal is probably not Andover, but the sequencing logic here cuts across every regulated industry. The instinct among most enterprise technology leaders right now is to run AI pilots in parallel with infrastructure cleanup, treating them as independent workstreams. Andover is explicitly rejecting that move, insisting that a governed, audit-ready data layer is a prerequisite, not a co-investment. If your organization is still treating data modernization and AI adoption as separate budget lines, this deal is a useful stress test of that posture.
The Medallion architecture choice is worth naming precisely because it signals something about Andover’s AI ambition. Medallion refers to a layered data organization approach, typically bronze for raw ingestion, silver for cleaned and conformed data, and gold for analytics-ready output, that creates a verifiable chain of custody from source system to AI model input. For an insurer subject to NYDFS scrutiny and AM Best rating pressure, that audit trail is not optional. The Cognizant Neuro AI platform sits on top of this, and its built-in explainability tooling is there to satisfy regulators who want to know why an underwriting recommendation was made, not just what it was. Any CIO in a regulated vertical who hasn’t yet asked their data team whether their current lake or warehouse can produce that kind of lineage on demand has a concrete question to bring to the next architecture review.
The most telling detail in this announcement is what Andover is explicitly not doing: replacing underwriter or adjuster judgment. Every AI use case named, submission intake acceleration, first notice of loss triage, insight surfacing, is positioned as decision support with a human in the loop. That framing is partly regulatory positioning, partly genuine risk management, and it matters because the alternative, autonomous decisioning, would trigger a different and far harder compliance conversation with NYDFS. If you’re designing AI workflows for any function that touches regulated outputs, the question isn’t whether to keep humans in the loop, it’s whether your vendor’s platform architecture actually enforces that constraint or just asserts it in the press release. Cognizant’s explainability and audit trail claims are worth pressure-testing in your next contract negotiation.
Based on reporting from The Andover Companies Selects Cognizant to Modernize Technology and Advance AI-Driven Innovation, originally published 2026-07-27 03:00:00.
